What is the difference between object detection and object recognition?

Instruction: Define both terms and explain the key differences.

Context: This question aims to evaluate the candidate's grasp of critical computer vision tasks and their ability to distinguish between closely related concepts.

Updated

Example Answer

Image classification labels a whole image. Object detection predicts where objects are and what classes they belong to, usually with boxes, class scores and confidence thresholds. Recognition can mean classification or identifying a particular instance, so I would clarify the task. For a shelf image, detecting three product boxes is different from classifying the image as a supermarket or identifying each exact product. A detector may use a CNN backbone; detection and CNNs are not competing categories.

Make it your own

Illustrative output contract: [{box: [x1,y1,x2,y2], class: "bottle", score: 0.91}]. Add instance IDs only if the task requires identification or tracking.

Why this works

The output contract makes the distinction concrete. A good whole-image label can coexist with missed small objects. Follow-up: how would you evaluate detection when boxes overlap?

Sources: TorchVision detection models.

Related Questions